Girls with large body size have a 15-30% reduced risk of developing breast cancer throughout life. This association is independent of both birth weight and adult body mass index (BMI) and is significantly stronger for estrogen receptor negative (ER-) disease. However, the underlying causes remain a mystery, especially since large body size at young age is correlated with earlier pubertal timing, a known risk factor for breast cancer. Self-reported childhood body size, a phenotype which is described in the literature as childhood body fatness, is a highly heritable trait. The correlation between childhood body fatness and adult BMI is only modest (r=0.24-0.20) and the estimated heritability of childhood body fatness is 70-80% compared to 40-70% for adult BMI. Still, the literature on genetic associations in pediatric body size-related phenotypes is very sparse and to date, no study has conducted a genome-wide search for loci associated with childhood body fatness. We here propose a genome-wide association study (GWAS) of childhood body fatness. We will use existing high-quality genotyped and imputed data for 2.5 million single nucleotide polymorphisms (SNPs) in 9,000 women in the Nurses' Health Study (NHS). As outcome, we will use recalled childhood body fatness averaged over ages 5 and 10 as assessed by a 9-level figure drawing. Previous validation studies have showed a high correlation between recalled body fatness and measured BMI (r=0.60 at age 5 and r=0.65 at age 10) indicating that these figure drawings provide an accurate assessment of body size at young ages. We will replicate the strongest associated SNPs in a Swedish population of 1,600 women (the SASBAC study), and three populations of children including 3,000 African-American children, 1,400 Hispanic children and 1,600 White non-Hispanic children. Confirmed variants will be tested for association with breast cancer risk using data from 4,000 breast cancer cases and 5,000 controls from nested case-control studies within NHS1, NHS2 and SASBAC and additional 2,100 ER- breast cancer cases from the NCI Breast and Prostate Cancer Cohort Consortium (BPC3). We will also investigate if the association with breast cancer differs with hormone receptor status. The rich resources in NHS and already completed genome-wide scans for 16,000 individuals provide excellent statistical power and a unique opportunity to study these critical questions in a highly cost-efficient manner. Identifying genetic predictors of childhood body fatness will provide invaluable insights into developmental and long-term processes that eventually affect breast cancer risk. Ultimately, untangling the complex etiology of breast cancer will help identify women at high risk as well as provide a platform for development of preventive and treatment strategies.